Curator: Efficient Indexing for Multi-Tenant Vector Databases
This addresses the need for efficient multi-tenant support in vector databases, enabling better resource sharing for applications handling unstructured data, and is incremental as it builds on existing indexing methods.
The paper tackles the problem of multi-tenancy in vector databases, where existing approaches trade off memory efficiency for search performance, and presents Curator, an in-memory index design that achieves low memory overhead and high performance for queries, insertion, and deletion, with evaluation showing search performance comparable to per-tenant indexing and memory consumption similar to shared indexing.
Vector databases have emerged as key enablers for bridging intelligent applications with unstructured data, providing generic search and management support for embedding vectors extracted from the raw unstructured data. As multiple data users can share the same database infrastructure, multi-tenancy support for vector databases is increasingly desirable. This hinges on an efficient filtered search operation, i.e., only querying the vectors accessible to a particular tenant. Multi-tenancy in vector databases is currently achieved by building either a single, shared index among all tenants, or a per-tenant index. The former optimizes for memory efficiency at the expense of search performance, while the latter does the opposite. Instead, this paper presents Curator, an in-memory vector index design tailored for multi-tenant queries that simultaneously achieves the two conflicting goals, low memory overhead and high performance for queries, vector insertion, and deletion. Curator indexes each tenant's vectors with a tenant-specific clustering tree and encodes these trees compactly as sub-trees of a shared clustering tree. Each tenant's clustering tree adapts dynamically to its unique vector distribution, while maintaining a low per-tenant memory footprint. Our evaluation, based on two widely used data sets, confirms that Curator delivers search performance on par with per-tenant indexing, while maintaining memory consumption at the same level as metadata filtering on a single, shared index.